Still, either they're using some private interface, or I can't see how they're doing format support checking, given that `IsFormatSupported()` is both documented and tested to always return false if channel count or sampling rate is different from the mix format, and it has no flags to alter its behavior.
I don't know, that obviously needs some exploration. But what if they don't use IsFormatSupported(), and instead just check things themselves and use audio client Initialize in a way which always supports necessary conversion?